Conducting market research

2.1. Who does the market research?Copy Link

Before you begin your market research, you need a plan to help identify the resources and timing of the research. Your organization needs to consider how much of the work will be done internally and how much will be outsourced.

The decision about who does the research will be a function of your specific needs, in-house capacity and financial constraints. Table 2.1 displays some options and the benefits and drawbacks of each one. You may choose to have different actors contribute to different parts of the process, or you may choose to have one actor take on the entire process. This Navigator can be a useful complement to any of the choices you make.

2.2. Setting hypotheses and developing research questionsCopy Link

A hypothesis is an informed guess: something you believe based on your prior experience, research or knowledge. For example, based on your own experience visiting clients with sales people, you may think that clients do not understand an insurance product. At this point, your opinion is only an informed guess, and you may want to ask clients additional questions to validate the hypothesis.

Alternatively, you may make an informed guess about something based on internal or external research and data analysis. For example, you may learn that few people in your target market own new vehicles. From that information, you might establish a hypothesis that there is a need for supplementary insurance for used vehicles.

If you host a design workshop as your first step, then your team will create hypotheses in the workshop. Box 2.1 provides more examples of hypotheses and demonstrates how to turn your hypotheses into research questions.

2.3. Identifying client segments, typologies or personasCopy Link

Developing inclusive insurance products requires truly knowing your current and potential clients. This includes demographic and financial characteristics, but also more abstract features, such as preferences, fears and values. Client segmentation is a very useful tool in market research, because it allows you to define the client types to focus on when you collect data. For example, if you see that a certain client segment is underserved, you may want to focus research efforts on that segment.

Segmentation allows you to group clients by their different needs. It will help group your target clients in ways that make the most sense for product design, distribution or servicing. Once you understand the specific needs of groups of clients, you can see the commercial benefits of customizing products, distribution and servicing a bit more to ensure every group is well served.

Even before you begin your research, you may have some ideas about the segments you want to target: for example, you might want to aim your product at rural women with loans from microfinance providers, or at urban bus drivers. A segment needs to be large enough to be attractive and have measurable characteristics that allow you to differentiate it from other segments. Some segments may be too broad, such as “bottom of the pyramid”, which would encompass many different types of people with different needs and financial capacities. Other segments may be too narrow: for example, “hairdressers” might contain too few people to justify a product. A measurable segment might instead be “urban businesses”, or “rural farmers”, or “factory workers”.

Gender is an important consideration when segmenting, as women and men often behave differently and have different needs resulting from household roles, social norms and economic inequities. But segmenting between men and women is not enough. Some women are wives and mothers, while some are single; some are employees and some have informal businesses, and so on.

Your client segmentation will only be as good as the data that you have, so having more data leads to more accurate client segments. Think about the segments and people you are planning to target. This will help you identify the internal and external data that you might need to deepen your understanding of these groups and validate that they are indeed an attractive market segment. It will also help you plan your data collection process if you are planning to conduct your own market research. Once you have some data, you can identify client segments using a range of methods, which may be as simple as descriptive tables or as advanced as machine learning clustering algorithms. If you lack data, consider the bottom-up segmentation technique explained in box 2.2.

After you identify your segments and understand them better through some research, you can assign these “typologies” using your data and understanding of the market. You may then want to personify these groups, creating “personas” that make them come alive within your team or when explaining their needs and behaviours to managers or potential partners.

Personas are fictional, but based on real information about clients. They can help bring to life, through description, an example of the type of person you are grouping into a segment. They may be a composite of different people you interview in a segment. Ideally, you should create personas with some research from the segment in hand, which will contribute to your understanding of the group more fully.

For example, Erica is a persona. She is not a real person. But in rural Indonesia (see case study 2.2), we met many women like Erica who worked in factories as employees. Many of them were offered multiple financial services, including direct deposit, through their work. Many travelled to work on motorcycles, and many had dreams of home ownership. From our conversations with these women, we could create a persona, whom we named Erica, that can help us imagine the person for whom we are designing when we start our product design process.

When you create personas, consider how gender disparities and social norms may influence the person and their perceptions of risk, financial vulnerability, knowledge and interest in insurance. Consider also whether some groups may have greater access than others to transportation, technology and information as you think about channels for enrolment, payments and claims. This will ensure that your persona is a reflection of your efforts to empathize with and understand your potential clients.

A final note: client segment development might be an iterative process. You might identify certain segments and then during external data collection find that they are more fluid than expected, or that you missed sub-segments. So, after you gather additional information, iterate and revisit your segment.

You may not always have access to adequate data sources to define segments. Bottom-up segmentation can validate hypotheses about market segments in your current market drawn from quantitative data or, in some instances, can entirely replace this information with insights from the people who have regular contact with your current or target clients. Box 2.2 offers guidance on conducting bottom-up segmentation in a focus group setting.

2.4. Selecting your data collection methodsCopy Link

Once you have broadly sketched your segments, your working group can use them to identify what information you are missing about your segments. You may be able to use existing data from internal or external sources to supply this information. Any remaining research gaps need to be filled with new data collection using quantitative or qualitative tools.

Collecting quantitative data will allow you to gather information from a larger sample and generalize your findings to the market segment. These data are numerical or statistical, and best suited to answer questions like “how much?”, “how many?” or “which one?” Qualitative data, on the other hand, are valuable because they include the kind of descriptive information that allows you to gain insight into your market’s underlying opinions, reasons and motivations. These data are best suited to answer “why” and “how” questions.

Table 2.4 can help you identify which methodology or type of tool to use, depending on your research questions and gaps.

Internal data sources

Internal data can be a valuable tool in designing a demand-driven product. If you or your distribution partner have worked with clients in any capacity, these clients may be a great target market to start with when offering inclusive insurance for the first time. You already have a trusted relationship with them, you likely already have some data on them that can help you design products, and you can reach out and speak to them to learn more.

In developed or saturated markets, designing products for a current client base can allow providers to take advantage of existing relationships and build brand loyalty without directly challenging first movers. It can also facilitate a smoother entry into a market that is underdeveloped, because your current client base is made up of the type of people who are willing to engage with a financial institution.

Internal data can come from an insurance company’s people and systems, or from those of your distribution partners (financial services provider, mobile network operator, etc.). Internal data can tell you things about your clients (age, gender, family characteristics, address, online/social media engagement, claims or loan records and payment or sales receipts, etc.) and about some of their risks and behaviours (mortality, repayment struggles, etc.). However, they cannot tell you very much about why your clients make the choices they do. If you only have resources to use internal data, you can make some hypotheses or guesses about your clients’ needs. But remember, these are only guesses until you test or validate them.

Internal data can be general or specific. You might have data that are not client-specific, such as the geographic areas in which loan officers work. Alternatively, you might have very detailed client data, such as yearly income or alternative credit scores. If you begin to develop client segments, typologies or personas before this step, these data can help you segment the market and refine ideas around the specific types of people you are targeting. Table 2.5 gives some examples of internal data sources and how you might be able to use them.

External data sources

External data are best suited to research questions about broad market characteristics and trends. The availability and quality of external data will vary by market. Some countries have very basic census data, while others carry out annual or even quarterly financial inclusion surveys. Regardless of the depth of data available, external data will give you a broad picture of the market and allow you to be strategic about whom to target with products.

If you began to develop client segments, typologies or personas before this step, external data can help you segment the market and refine ideas around the specific types of people you are targeting. For example, you might decide not to develop a product for rural youth if external data show that youth are moving into urban areas. Or, you might discover that, in the specific geographic area of the country in which you want to launch your product, there are too few women like Erica, and thus the market is too small to ensure a profitable outcome.

2.5. Collecting new data – Conducting field researchCopy Link

Since truly inclusive insurance requires a client-centred approach, collecting new data can help you ensure that clients’ voices are heard throughout your process. Every situation is different and access to client voices can be very easy or very difficult to obtain, depending on your organization’s contact and engagement with end users.

When thinking about collecting new data, consider your closest points of contact to your target clients. Do you have a distribution partner with close ties to the community? Do you have active social media platforms that engage your target clients? Maybe your front-line staff know your target market well and can offer some interesting insights, or maybe you have access to others’ front-line staff?

This section can help you think through how to collect data, how to decide on the types of data you want to collect and how to develop the tools that your team will use to collect the data. Alongside the guidance in this section, you will also find several useful resources in the Toolkit.

Remember that once you reach your target market with questions, you need to leverage this point of contact for as many types of information as you can: not only on product needs but also on pricing, marketing, distribution and payment channels, and more. This does not mean you need to have lengthy and time-consuming questionnaires – only that you need to consider your top priority questions from the start, and make sure they are covered in your research.

2.5.1. Quantitative data collection tools

Quantitative data are often collected through surveys, which allow the researcher to ask a large number of people the same questions to reflect market trends. Collecting this kind of data can help you to find out representative information about a group or segment.

In quantitative data collection, closed-ended research questions are best. For example, you might ask which financial products a person owns, how many people they support financially, whether they bank online, or how much money they have in savings. You might provide multiple choice or binary options as potential answers to these questions.

The first step in developing data collection tools is to turn your research questions into data collection questions. Data collection questions can be closed-ended (with specific options or choices) or open-ended (which can be answered in any way that the person being questioned wishes), and these question types are suited to different methodologies: closed-ended questions are more often used in quantitative data collection, while open-ended questions lend themselves to qualitative data collection. Either way, to facilitate analysis, the questions need to be consistent.

The questions should also be phrased simply, using language that is understandable to the respondent (the person who is the subject of the research). By pushing yourself to use simple language, you will be moving one step closer to your future marketing strategy. You will start to test what is “simple” and “easy to understand” for your potential customers. Box 2.5 gives you some examples.

Box 2.6 considers the sample hypotheses and research questions from the themes in box 2.1 . It illustrates how to turn a hypothesis into a research question, and then how to make these research questions easily understandable for your audience by turning them into data collection questions.

The exercise also explains the methodology that might be best to use to collect the information from your respondents. When you begin to discuss your questions with clients, there is no need to start by saying that you are developing an insurance product. Instead, you should explain you are looking to offer services that will help them manage costly events. If they have a negative image of insurance, mentioning it at the outset might influence their attitudes about your conversation and colour the responses you get.

Remember that data collection takes time and that respondents become tired after they have spent time answering questions. To ensure that you keep your participants’ attention and so get thoughtful and honest answers, a guiding rule is that closed-ended quantitative surveys should take no longer than 15 minutes over the phone or 20 minutes in person. Individual qualitative interviews can take longer, up to 30 minutes, because they are more engaging and conversational. Focus group discussions should be no longer than 75 minutes. Also remember that more data is not always useful; in fact, too much data can make the data analysis process more difficult to navigate.

Box 2.7 discusses how to create surveys for inclusive insurance, and the Toolkit contains several other resources to help you author surveys.

Once you have decided on your survey instrument, you need to decide whether the information will be collected on paper or digitally. Paper surveys and questionnaires are more operationally complex, because they need to be printed, numbered, digitized and stored. However, if surveyors do not have access to mobile phones or tablets, paper surveys can be less expensive than digital alternatives.

In the past, digital instruments relied on Wi-Fi or data connectivity on the ground, which could be sparse and unstable. Today, free tools such as KoBoToolbox are designed to work offline with most smartphones, tablets or desktop computers. Digital instruments can be more efficient, because you only need to enter data once, so information does not need to be re-typed. However, you should put in place protocols for uploading and reviewing quality to ensure that surveyors are accurately following the process. Protocols should include rules for how respondents’ data privacy will be secured, aligned with local regulatory privacy norms and regulations.

2.5.2 Qualitative data collection tools

Think about qualitative data collection as a series of guided conversations. These conversations can be individual conversations, conducted through interviews, or group conversations, conducted through focus groups. Research questions best suited to qualitative data collection are exploratory and open-ended. For example, you might ask questions about what financial risks someone faces, what financial anxieties someone has, or how someone prioritizes their expenses.

• Interviews allow you to speak with a single person in great depth and to ask them questions about sensitive topics that might not be appropriate in a group setting. Interviews also enable you to record individual stories, which can be a powerful tool for understanding your market.
• Focus groups allow respondents to engage in conversations with each other (guided by your data collection questions). Focus groups are a useful format for listing and ranking client needs and preferences, since participants can work together to agree on lists and ranks that resonate with them as a group. They are also the best medium for customer journey mapping, and for gaining insight into how a market segment discusses its needs and preferences – your team will want to mirror this language in marketing materials.
• It is important to conduct focus groups with each client segment separately, both to learn as much as possible about that segment and to create a comfortable environment for participants. For example, having Erica and her fellow garment workers in one group would allow you to gather more information about how they are paid, the financial instruments they use, and their hopes and dreams for the future. This can help you consider which assets female garment workers are most interested in protecting. Box 2.8 provides guidance on conducting focus groups.
• Mystery shopping is typically seen as supply-side research, enabling you to find out what your competitors are offering, as well as how and through which channels they are offering it. Mystery shopping can also provide a more nuanced understanding of customers’ experience during the sales process. When you or your team put yourselves in the customers’ shoes, you can feel whether customers are being treated respectfully, observe whether customers are receiving consistent and clear messages about the product and find out how staff respond to questions. The Consultative Group to Assist the Poor (CGAP) offers a Guide to Mystery Shopping for financial services that can help orient your team on this path.

Most qualitative analysis involves small samples, because it does not seek to be representative. This means data can be analysed using spreadsheet software such as Microsoft Excel or any simple matrix format. Specialized software is also available for larger and more complex data sets. This kind of software can also be useful when organizational capacity for sophisticated analysis is low. Our Toolkit offers resources that can help.

2.5.3. Sampling and recruitment strategies

Your research sample is made up of the people whom you will survey, interview or speak with in a focus group. Ensuring that you choose the right people for your research is as important as designing your research tools. They have to be comfortable speaking to you and not be worried that their jobs, loans or commercial relationships might be affected by what they say. It is important to invite people as randomly as possible to avoid only selecting those who are most engaged and outspoken.

You also need to ensure that when people are speaking to you, they are in a space that feels safe and private. In practice, this may be difficult. Some people from your segments might live very far away and lack access to transportation to reach you. Some people may be uncomfortable participating. To ensure you hear from a broad range of people, you might consider having a trusted person or institution invite respondents, you may need to travel to respondents’ communities, and you might offer participants a small incentive.

To begin, think about the kinds of people you want to have in your research sample. If you have started to create segments before this step, you can use them to determine the right people to speak to, as well as to ensure that you interview enough members of each segment or typology. If you have a large pool of clients to draw from and they represent the market well, you can focus on recruiting your clients for the research. If you have identified client segments that are not represented in your client base, you will want to include outside respondents. For example, if your market includes a high frequency of college-educated, single women, but you currently have no products that serve this segment, you may want to make a connection with a local university to recruit female alumni.

Another important decision is defining the number of people to whom you will speak. This is your sample size. Different methodologies require different sample sizes and determining a sample size may be partially dictated by your capacity to reach people and by your budget.

When sampling respondents for qualitative data, you should aim for 13–20 respondents per segment. After about 10 respondents, you begin to see patterns you can analyse in preferences and behaviours. For focus groups, two groups of 6–8 people per segment are better than one large group. This way, a group dynamic that goes awry can be mended by a second, similar population and findings can be validated between groups. Qualitative data is not statistically significant or representative. You are just looking to discuss the answers to the questions “why?” and “how?” and to understand them across different experiences.

Selecting the right sample size is a critical step in quantitative data gathering. Tools like SurveyMonkey and Creative Research Systems offer guided processes for determining sample sizes, and generative artificial intelligence (AI) tools, such as ChatGPT, Claude, Copilot or Deepseek, can also help with this process if you provide specific inputs (see box 2.10 for an example). Inputs might include the total size of your population or market, your hypotheses about the differences you expect to observe between groups or segments, the desired confidence level for your results, and the acceptable margin of error. For instance, a 95 percent confidence level and a margin of error of ±5 percent are standard parameters in many studies.

If your goal is to assess statistical significance, you will need to account for statistical power in your sample size calculations. Statistical power ensures you have a high probability of detecting meaningful differences in your data, typically set at 80 percent or higher. Many online sample size calculators automatically include this consideration.

However, if your focus is on exploratory research to gain insights or describe your market, and you are constrained by time or budget, statistical power may not be your priority. In these cases, a practical guideline is to aim for at least 200 total survey responses, with about 60 responses per segment for meaningful subgroup analysis. Of course, more is always better, but less is workable too – just try to get as many respondents as you can! Finally, if you are conducting an academic survey, you will likely want to work with an academic researcher or institution.

2.5.4. Best practices: Respecting participants and data protection

Whether you are speaking with your clients or with other members of the population, be mindful of some important principles in collecting this type of data, as outlined in box 2.11.

2.5.5. Planning fieldwork for data collection

When your data collection instruments and sample recruitment plans have been finalized, your team will need to finalize a data collection plan. Along with the guidance in this section, the Toolkit contains resources for managing data collection, including tools that can be used to ensure gender inclusivity is considered throughout this process.

Data collection includes managing the fieldwork, including the recruitment, training and supervision of the people conducting the data collection, as well as managing the data that are collected to ensure they are complete and well organized. Fieldworkers perform the researcher tasks of surveying, interviewing and facilitating focus groups. Make sure that your fieldworkers are experienced with the target population, and ideally ensure a gender balance to enable people of all genders to speak more freely and comfortably.

Regardless of who is doing the fieldwork (whether it is surveyors from a survey firm, loan officers from a distribution partner, etc.), training should be provided to ensure that questions are asked in the same way and with the same intention to all respondents. This is an important part of the research process that can be easy to let slide, but without providing formal training, you risk wasting time, funds and energy collecting data that do not adequately answer your research questions. Box 2.12 provides tips on recruiting, training and compensating surveyors, interviewers and focus group facilitators.

Training should include a session on how to ensure that you are respectful and inclusive of all types of respondents. This includes women, youth, the elderly and people with disabilities. It might be helpful to have a discussion to explore the biases that fieldworkers may have when recruiting and talking to respondents, to ensure that they are aware of the need to be as inclusive as possible. Fieldworkers may come from different socioeconomic groups or age groups and may not be sensitive to the types of constraints older or more vulnerable respondents face. Fieldworkers should also be instructed to be gender inclusive, highlighting that in some cultures and contexts, women may prefer to be interviewed separately from men. In other contexts, women may need to be spoken to with more caution, since they may have reservations about revealing personal information. Consider the perspective of your more vulnerable respondents and ensure they are reflected in your approach.

You will also need to make sure you have a plan for quality checks. Checking surveyors, for example, could involve calling a few random respondents and ensuring that they spoke with the surveyor, or it could include analysing data for outliers or patterns among surveyors that might indicate they are making up (fake) data. To quality check interviews and focus groups, you might listen to parts from a few recordings.

Unless you are outsourcing all of the fieldwork to a market research firm that has protocols in place, you will want to think about how to best compensate and incentivize fieldworkers. Consider paying a small daily rate for “showing up”, alongside a fee per completed survey for surveyors (and make sure these are reviewed before payment), or for interviewers and facilitators, a fee per completed interview/focus group and relevant notes. Reimburse fieldworkers for transportation and airtime as well as any other work-related expenses.

Data collection involves determining the form in which you want the data. For surveys, this might mean a plan for downloading data from mobile devices, or a programme for data entry when paper surveys are being used. For interviews and focus groups, you will need to think about the requirements of your analysis plan. Do you want interviews and focus groups recorded? Do you want interviewers and facilitators to take notes? Would it be useful to have recordings transcribed (written down word for word)? Are you comfortable using AI tools for any of these functions and if so, what privacy protocols should you follow?

2.6. Customer journey mappingCopy Link

Customer journey mapping is a qualitative tool that relies on understanding and empathizing with the client around their entire experience with a product. It can provide the key to discovering frictions in the customer journey as clients learn about how to acquire, pay for and make claims on an insurance product. This can improve the customer experience and ensure greater acquisition, payment compliance and claims.

Remember that inclusive insurance needs to offer value for end customers, so it is essential that customers understand products and make claims when they are due. That means that processes need to be clear and simple, with as little friction as possible. The customer journey map can help identify which products, services, actions and interactions or touchpoints will lead to a smooth customer experience.

You can map your customer journey by individually interviewing a sample of clients about their experiences, or by conducting focus group discussions. Before you begin, it is a good idea to capture in which segment each participant belongs, since different types of clients may have very different perceptions and experiences, giving them very dissimilar journeys. You can reference the qualitative data collection section to learn more about preparing for and running client interviews and focus groups.

Box 2.13 discusses how to lead an individual or group in a customer journey mapping exercise. Customer journey mapping can also be conducted during design workshops to map out the process that your institution would ideally imagine for a client.

2.7. Analysing your dataCopy Link

Most projects suffer not from too little data, but from too much data. Keeping your data collection tools short and to the point will reduce the complexity of data analysis and make it more effective. To be as efficient as possible, think about your analysis plan before you create your data collection tools. Remember that your analysis needs to be unbiased, and that it needs to consider the results of the research rather than your initial hypotheses and opinions.

Analysis of qualitative interviews should focus on how respondents (by segment) manage risks and their perceptions of the difficulties of their risk management tools as well as of the likelihood of the events that pose risk. Analysis should also assess which delivery channels are most trusted.

You might find qualitative software programs such as Dedoose or Nvivo helpful in analysing your data. This kind of software is particularly useful if you have large sets of data or if you do not have analytical personnel available to work on the study. You can also use a spreadsheet table, using Microsoft Excel or another spreadsheet application, with questions as the column headings and respondents as the rows. This can enable you to compare responses across individuals and identify patterns. Remember to make note of the demographic characteristics of respondents to ensure that your analysis includes segmentation. Focus group analysis can also be grouped in tables, with one row for each group.

Finally, generative AI tools can support you in the process of data analysis. You can start by preparing your transcripts to ensure they are clear and organized, removing unnecessary filler words while preserving the original meaning. Then, you can provide an AI tool such as ChatGPT with some brief context for the project, such as explaining that the focus group discussed barriers to accessing inclusive insurance. Then, share a portion of the transcript and ask for specific insights, such as identifying common themes, highlighting challenges, extracting key quotes or summarizing participants’ recommendations. For example, you might say, “This is a transcript from a focus group on inclusive insurance. Please identify the main themes and barriers discussed and highlight any suggestions or participant quotes about improving access.” The AI tool will analyse the text and return a summary or analysis, which you can refine further by asking follow-up questions tailored to your specific focus.

This process can help you quickly uncover meaningful insights to guide your work. However, be aware that it is not a perfect solution. Part of the “magic” in analysing voices from clients is in understanding nuances that transcripts may miss. Consider people’s body language, moods and tones during discussions. Were they relaxed? Did they fully understand the topic? Did they suggest that some of their verbal responses were too optimistic, perhaps to please you or a partner organization? It is important to make sure that technology is your helper and not a crutch.

Analysis of quantitative data should begin with thorough data cleaning. Before you begin, go through your data and “clean” them. First, eliminate duplicate and blank entries. After this, consider coding some open-ended answers or adjusting any inconsistencies: for example, if the answer to a question is “3”, and the responses are written as “3”, “three” and “III”, these should all be recoded to “3”. The analysis can be conducted in a statistical software package or in a spreadsheet application such as Microsoft Excel, depending on internal capacity.

You can learn a lot from survey data without doing anything sophisticated. Begin with basic descriptive tables that show who was interviewed (gender, age, occupation, income, etc.), as well as tables of survey responses to each question, giving frequencies and averages depending on the questions. If you want to know how many people answered that they have a loan, you can show the frequency of this response (e.g., 40 percent). If you want to know how often people see a doctor, you may want to show an average (mean) and standard deviations if you have a large data set. You can then make tables of the results by segment to explore the risks, preferences and financial constraints specific to each segment.